AI Security: Intel & Dell Protect Against Prompt Injection

Key Takeaways

• Experts from Dell and Intel are securing AI infrastructure from the ground up to protect against threats like prompt injection.
• Insider threats now include employees using unapproved chatbots or AI agents with excessive permission to access sensitive data.
• Agentic AI systems have moved from research to live production environments where they plan and execute tasks with minimal human help.
• Recent incidents involving Anthropic show how configuration errors can let AI models access real-world systems unexpectedly.
• Alpha Vision will display AI security tools at the Global Security Exchange in Atlanta from September 14 to 16 at Booth #3050.
• The Alamance-Burlington school board is testing an AI security system using facial recognition to spot potential threats in schools.
• Mistral has partnered with Cloudera to integrate models that help businesses own their data and intelligence within their own environments.
• Michigan State University President Kevin Guskiewicz aims to make the school a leader in AI while keeping human values at the center.
• Author JG Christman argues that the main issue with AI is whether humans can still make their own judgments and take responsibility.
• Education systems need to teach students how to collaborate with AI and recognize when it makes mistakes or hallucinates.

AI Factories Create New Security Risks for Enterprises

Companies are building AI factories to produce knowledge at scale, but these systems introduce new security risks. Traditional cybersecurity models cannot handle data that is continuously generated and consumed by autonomous AI agents. Experts from Dell and Intel are now securing AI infrastructure from the ground up to protect against threats like prompt injection. Security leaders worry that agents acting without human review could take harmful actions in real time.

Insider Threats Evolve With AI and Unchecked Access

The definition of an insider threat is changing as organizations adopt generative AI and agentic systems. Risks now include employees using unapproved chatbots or AI agents with too much permission to access sensitive data. Experts like Bertijn Eldering and Clyde Williamson warn that legitimate access can expose more information than intended. Companies must give agents specific identities, limit their tasks, and keep humans in the loop for important decisions.

Agentic AI Brings Power and New Security Challenges

Agentic AI systems have moved quickly from research to live production environments where they plan and execute tasks with minimal human help. These autonomous agents can use tools and APIs to achieve goals, but they also create new points of failure for security teams. Securing these agents requires strict identity management and monitoring of their reasoning processes and memory stores. While they help defend against threats, their independence expands the potential for errors or malicious actions.

AI Agents May Become the Next Customer Experience Risk

Enterprise AI agents are gaining access to systems that manage customer experiences, creating new risks for business leaders. Recent incidents involving Anthropic and other companies show how configuration errors can let AI models access real-world systems unexpectedly. Security teams are now deploying their own AI agents to defend these environments, which requires careful control over permissions. Both customer-facing and security-facing agents need clear boundaries to prevent harm to customers and data.

Gen Z Needs AI Fluency Instead of Just Banning Tools

Schools are currently restricting AI use, but this approach leaves students unprepared for future jobs where AI is widely adopted. Hiring managers look for candidates who can use AI to draft and research while still thinking critically about problems. Education systems need to teach students how to collaborate with AI and recognize when it makes mistakes or hallucinates. The goal is to balance tool usage with the critical thinking skills that make workers valuable.

Alpha Vision shows AI security tools at GSX 2026

Alpha Vision will display its new AI security tools at the Global Security Exchange in Atlanta from September 14 to 16. The company will be at Booth #3050 to show how its AI Agents turn regular cameras into smart security resources. Visitors can try features like Magic Search to find people or objects quickly and use Virtual Guard for real-time voice commands. New updates include tracking people across multiple cameras and automatically checking if workers wear safety gear like hard hats. These tools help security teams review video faster and create reports without doing repetitive manual work.

Burlington schools test AI system for safety

The Alamance-Burlington school board in North Carolina recently reviewed a new AI security system designed to protect students. The system uses facial recognition and machine learning to spot potential threats like someone entering a restricted area. It is currently being tested in several schools to see if it can prevent accidents and keep students safe. The board was shown a demonstration that included a simulated accident to show how the system responds. Officials stated the system will be used only for safety and will not be used to monitor student behavior or track individuals.

Mistral and Cloudera partner for sovereign AI

French AI startup Mistral has partnered with American company Cloudera to meet the growing demand for sovereign AI. This type of AI keeps data, intelligence, and operations under the customer's control within their own environment. The two companies will integrate Mistral's models with Cloudera's platform to help businesses own their data and intelligence. Ricky, a spokesperson, explained that this shift moves companies from renting generic AI to owning intelligence tuned to their specific needs. The partnership allows customers to use their own proprietary data to train models that work inside their own systems.

MSU President aims for AI leadership with ethics

Michigan State University President Kevin Guskiewicz wants the school to become a leader in artificial intelligence. He recently met with students, faculty, and alumni to discuss using AI responsibly and ethically. The group agreed that while the university should embrace AI, it must keep human values at the center of its work. Guskiewicz plans to add more instruction on AI for incoming students and create systematic guidance for its use. He emphasized that the university wants to build an AI-ready workforce while ensuring the human factor remains the most important part of any decision.

New book explores AI and human moral choice

Author JG Christman released a new book titled Good and Evil in the Age of Simulation that examines the relationship between AI and human morality. The book argues that the main issue is not how smart AI becomes, but whether humans can still make their own judgments and take responsibility. Christman looks at ancient moral traditions and modern technology to explain how people should handle powerful systems that try to make decisions for them. He warns against letting efficiency or safety goals replace truth and human freedom. The book encourages readers to maintain their ability to judge and refuse when systems try to control their choices.

Sources

• AI attack surface reshapes enterprise security
• Insider Threat Awareness Month: Rethinking Risk in the Age of AI
• The Role of Agentic AI in Cybersecurity
• The Next CX Security Risk May Be the AI Agent You Gave Access
• These are the AI skills Gen Z should have
• Alpha Vision to Showcase AI-Powered Physical Security Intelligence at GSX 2026
• School board learns about AI security system
• Mistral and Cloudera Team to Meet Sovereign AI Demand
• Guskiewicz wants MSU to be a leader in AI
• Good and Evil in the Age of Simulation Confronts Artificial Intelligence, Moral Power and Fight to Preserve